Upload app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
from dotenv import load_dotenv
|
| 3 |
+
import os
|
| 4 |
+
from langchain_community.llms import HuggingFaceEndpoint
|
| 5 |
+
|
| 6 |
+
# try:
|
| 7 |
+
# load_dotenv('C:\\Users\\H587077\\Desktop\\study\\temp\\api-keys.txt')
|
| 8 |
+
# except:
|
| 9 |
+
# print("Environment file not found !! MUST find the env var HUGGINGFACEHUB_API_TOKEN to work.")
|
| 10 |
+
|
| 11 |
+
# Title
|
| 12 |
+
st.title('HuggingFace LLM playground')
|
| 13 |
+
|
| 14 |
+
models = [
|
| 15 |
+
'mistralai/Mistral-7B-Instruct-v0.2',
|
| 16 |
+
'google/flan-t5-xxl',
|
| 17 |
+
'tiiuae/falcon-40b-instruct',
|
| 18 |
+
]
|
| 19 |
+
|
| 20 |
+
# Selected model in model_id
|
| 21 |
+
model_id = st.sidebar.selectbox(
|
| 22 |
+
'Select model',
|
| 23 |
+
options=tuple(models)
|
| 24 |
+
)
|
| 25 |
+
|
| 26 |
+
if 'model-response' not in st.session_state:
|
| 27 |
+
st.session_state['model-response'] = '<provide query & click on invoke>'
|
| 28 |
+
|
| 29 |
+
# draw the box for model response
|
| 30 |
+
st.text_area('Response', value = st.session_state['model-response'], height=400)
|
| 31 |
+
|
| 32 |
+
# draw the box for query
|
| 33 |
+
query = st.text_area('Query', placeholder='provide query & invoke', value='who was the president of the USA in 2023?')
|
| 34 |
+
|
| 35 |
+
# Temperature
|
| 36 |
+
temperature = st.sidebar.slider(
|
| 37 |
+
label='Temperature',
|
| 38 |
+
min_value=0.01,
|
| 39 |
+
max_value=1.0
|
| 40 |
+
)
|
| 41 |
+
|
| 42 |
+
# Top p
|
| 43 |
+
top_p = st.sidebar.slider(
|
| 44 |
+
label='Top p',
|
| 45 |
+
min_value=0.01,
|
| 46 |
+
max_value=1.0,
|
| 47 |
+
value=0.01
|
| 48 |
+
)
|
| 49 |
+
|
| 50 |
+
# Top k
|
| 51 |
+
top_k = st.sidebar.slider(
|
| 52 |
+
label='Top k',
|
| 53 |
+
min_value=1,
|
| 54 |
+
max_value=50,
|
| 55 |
+
value=10
|
| 56 |
+
)
|
| 57 |
+
|
| 58 |
+
repetition_penalty = st.sidebar.slider(
|
| 59 |
+
label='Repeatition penalty',
|
| 60 |
+
min_value=0.0,
|
| 61 |
+
max_value=5.0,
|
| 62 |
+
value=1.0
|
| 63 |
+
)
|
| 64 |
+
|
| 65 |
+
# Maximum token
|
| 66 |
+
max_tokens = st.sidebar.number_input(
|
| 67 |
+
label='Max tokens',
|
| 68 |
+
value=50
|
| 69 |
+
)
|
| 70 |
+
|
| 71 |
+
# Function to create the LLM
|
| 72 |
+
def get_llm(model_id):
|
| 73 |
+
return HuggingFaceEndpoint(
|
| 74 |
+
repo_id=model_id,
|
| 75 |
+
temperature=temperature,
|
| 76 |
+
top_k = top_k,
|
| 77 |
+
top_p = top_p,
|
| 78 |
+
repetition_penalty = repetition_penalty,
|
| 79 |
+
max_new_tokens=max_tokens
|
| 80 |
+
)
|
| 81 |
+
|
| 82 |
+
# Function for invoking the LLM
|
| 83 |
+
def invoke():
|
| 84 |
+
llm_hf = get_llm(model_id)
|
| 85 |
+
|
| 86 |
+
# Show spinner, while we are waiting for the response
|
| 87 |
+
with st.spinner('Invoking LLM ... '):
|
| 88 |
+
st.session_state['model-response'] = llm_hf.invoke(query)
|
| 89 |
+
|
| 90 |
+
st.button("Invoke", on_click=invoke)
|